Question

I’ve been trying to figure out an efficient way to work without reading twice from a file when using a buffer to search for a pattern of bytes. I’ve chosen to implement Runnable so I can divide my task to work in concurrent threads. My code looks something like this:

// constructor: initializes local variables.
public BytePatternSearcher(RandomAccessFile raFile, byte[] pattern, int bufferSize, long startPos, long endPos);

public void run()
{
    while(amountToRead - raFile.read(buffer) > 0)
    {
        // search code
    }
{

Now, I’ve hit a snag: my algorithm works in simple cases, but not in complex ones. I made the assumption that there are no cases of a pattern starting within one already being searched, that the pattern length is shorter than the buffer, and so on, limiting to one scan at a time and just iterating through the file. Naturally, this is not a very robust solution. Suppose I have a pattern of ‘xxxxx’ (length 5), my file is ‘xxxxxxyxxxxxx’ and my buffer size is 2 (x and y represent certain byte values). The string appears 4 times, and each check requires more than twice the buffer length.

How do I go about working things out without checking the same byte more than once for all cases?

Answer 1

The wikipedia has an entry for Boyer–Moore string search algorithm which also contains some sample implementations.